Download Lessons from Fair Lending Law for Fair Marketing and Big Data

LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
Peter Swire
Nancy J. & Lawrence P. Huang Professor of Law and Ethics,
Georgia Institute of Technology
Senior Fellow, Future of Privacy Forum*
September 11, 2014
Executive Summary
This White Paper is submitted in connection with Peter Swire’s participation at the
Federal Trade Commission’s September 15, 2014 event on “Big Data: A Tool for
Inclusion or Exclusion?”
Issues of potential discrimination in online marketing have recently attracted increasing
attention, notably in the April 2014 White House Report on Big Data and Privacy. To
date, however, the public debate has not included a systematic analysis of the closest
legal precedents – fair lending law under the Equal Credit Opportunity Act (ECOA) and
other statutes. This White Paper, drawing on my previous scholarship in the fair lending
area, 1 explains lessons from fair lending law for what I call “fair marketing” – the
application of anti-discrimination and fair lending principles to online advertising and the
Big Data analytics that support much of online advertising.
This White Paper applies fair lending law in two principle ways:
1. Sectoral legislation applies to some online marketing, much as sectoral
legislation for privacy exists in areas such as HIPAA, the Gramm-LeachBliley Act, and the Children’s Online Privacy Protection Act. Antidiscrimination law applies to marketing for lending, housing, and employment,
under the ECOA, the Fair Housing Act, and Title VII of the Civil Rights Act of
1964. An important legal topic going forward will be whether and in what ways
these sectoral anti-discrimination laws apply to entities in the online advertising
ecosystem who play a role in advertising for lending, housing, and employment
as well as other sectors.
*
The opinion expressed herein are those of the author alone, and do not necessarily reflect the
view of participants in the Future of Privacy Forum. Research assistance provided by Joseph
Jerome, Policy Counsel, Future of Privacy Forum.
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
2. Over two decades of fair lending enforcement and regulatory guidance
provide numerous useful insights for advertising in sectors other than
lending, housing, and employment, where sectoral statutes do not
currently exist. Financial institutions have extensive and long-developed
programs for fair lending compliance. Instead of fair marketing issues being
entirely novel, these compliance programs provide well-established mechanisms
for addressing possible discrimination in marketing.
Part I addresses the law and history of fair lending enforcement. Fair lending laws (which
historically have included fair housing) have broad scope and application. The ECOA, for
example, prohibits discrimination with respect “to any aspect of a credit transaction . . .
on the basis of race, color, religion, national origin, sex or marital status, or age . . . .” It
defines “creditor” broadly to include “any person who regularly extends, renews, or
continues credit,” as well as brokers and “any person who regularly arranges for the
extension, renewal, or continuation of credit.” The Fair Housing Act and Title VII have
similarly broad language with respect to housing and employment discrimination.
The disparate impact theory has long played a central role for enforcement of fair
lending laws. A 1994 bank regulator policy statement defined disparate impact as “when
a lender applies a practice uniformly to all applicants but the practice has a
discriminatory effect on a prohibited basis and is not justified by business necessity.”
Fair lending enforcement became active during the 1990’s, and received renewed
attention in response to subprime loans and the financial crisis.
Part II examines important possible distinctions between fair lending and fair marketing,
to determine whether and what ways online marketing should be treated differently from
fair lending. Specifically, I assess: (1) financial services are heavily regulated and
supervised, in contrast to online marketing; (2) financial services are dominated by large
companies, in contrast to the start-ups and small businesses engaged in online
marketing; (3) whether congressional action in fair lending suggests other areas should
be less regulated; (4) whether a history of government action perpetuating discrimination
would be relevant to fair marketing; and (5) whether anti-discrimination principles should
apply only to a limited set of transactions of highest importance to consumers. Upon
examination, I do not discern any categorical reason why online marketing should be
exempt from the anti-discrimination principles embodied in the ECOA, Fair Housing Act,
Title VII and numerous other laws.
Part III offers four lessons from fair lending for fair marketing:
1. Fair lending law affirmatively encourages targeted marketing to protected
classes. In contrast to the skepticism of targeted marketing voiced by some in
the privacy debates, fair lending consent decrees and regulatory guidance have
often encouraged or required such marketing to members of historically underserved communities.
2. Fair lending law applies disparate impact analysis to targeted marketing.
Financial regulators also see risks in targeted marketing, as exemplified by
subprime mortgage loans. As applied to fair marketing, the experience with fair
lending shows an apparent paradox – targeted advertising to reach minority
groups and protected classes is often expected, yet it is illegal to use targeted
ii
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
marketing to steer products to those groups. In my view, this topic needs
considerably more research and discussion.
3. Sectoral statutes prohibit discrimination in online marketing for lending,
housing, and employment. A crucial legal fact about fair marketing is that the
ECOA, Fair Housing Act, and Title VII apply to both online and offline marketing.
This White Paper identifies a number of issues under these statutes that require
additional attention, to determine the impact of existing law on online marketing
practices.
4. Data relevant to a disparate impact analysis may often be more available
for online marketing than for traditional fair lending cases. The history of fair
lending enforcement has been intertwined with the limited availability of data to
detect and prove possible violations in anti-discrimination laws. These data
limitations may be easier to address in many online advertising situations, which
are characterized by precisely the Big Data that is the subject of this FTC
Workshop.†
†
I also add a disclaimer to the analysis in this White Paper. I am presenting this document while
in the midst of research for what I expect to be a longer research effort on fair lending and fair
marketing. This White Paper was completed under the time constraints of preparing for the FTC
workshop, and I have not yet had the benefit of comments on this written product. The application
of anti-discrimination law to online marketing presents complex issues, and so I present here my
best current thinking but with an awareness that my views may shift as the research project
continues.
iii
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
I. The Law and History of Fair Lending Enforcement
This part of the White Paper first explains the provisions of the Equal Credit Opportunity Act that
are most relevant to fair marketing. It briefly examines other statutes that have been important
to the development of fair lending law and anti-discrimination law generally, including the
Community Reinvestment Act of 1977, the Fair Housing Act of 1968, and Title VII of the Civil
Rights Act of 1964. I then examine the relevant history of fair lending enforcement.
A.
The Equal Credit Opportunity Act
The Equal Credit Opportunity Act was enacted in 1974. An important original purpose was to
ensure that married and divorced women could establish credit in their own names. The statute
applies far more broadly than that, however. It makes it “unlawful for any creditor to discriminate
against any application, with respect to any aspect of a credit transaction – (1) on the basis of
race, color, religion, national origin, sex or marital status, or age (provided the applicant has the
capacity to contract.”2
Exceptions provide that creditors can take marital status or age into account for limited nondiscriminatory reasons, such as to ascertain a creditor’s remedies or to “use any empirically
derived credit system which considers age if such system is demonstrably and statistically
sound” in compliance with the ECOA regulations.3 The Consumer Financial Protection Bureau
(CFPB) now issues these regulations, known as Regulation B.4
The definition of “creditor” is broad, including “any person who regularly extends, renews, or
continues credit,” as well as brokers and “any person who regularly arranges for the extension,
renewal, or continuation of credit.”5
Enforcement of the ECOA occurs through the CFPB and the federal banking agencies, such as
the FDIC and the Federal Reserve. In addition to specific ECOA actions, each of these
agencies can use its existing supervisory and enforcement powers to ensure compliance with
the ECOA.6
The statute specifically provides that the Federal Trade Commission is authorized to enforce the
requirements of the ECOA.7 A violation of the ECOA is deemed a violation of the Federal Trade
Commission Act, and the FTC can use all of its functions and powers under that Act to enforce
compliance by any person who violates the ECOA’s requirements. In contrast to general
limitations on the FTC’s jurisdiction to persons engaged in commerce, the statute provides that
the FTC can enforce compliance with the ECOA “irrespective of whether that person is engaged
in commerce or meets any other jurisdictional tests under the Federal Trade Commission Act.”8
The FTC can enforce any rule prescribed by the CFPB “in the same manner as if the violation
had been a violation of a Federal Trade Commission trade regulation rule.”9 The FTC cannot
enforce against entities supervised by the FDIC, the Federal Reserve, and other listed federal
agencies.10 It can, however, enforce the ECOA for entities that are also subject to enforcement
by the CFPB.
The statute also provides for civil liability. The Department of Justice may bring a civil suit when
there is a “pattern or practice” of discrimination, including for actual and punitive damages and
injunctive relief.11 Aggrieved individuals can bring individual actions, with actual damages plus
1
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
punitive damages up to $10,000. Class actions are also available, with punitive damages up to
the lesser of $500,000 or one per cent of the creditor’s net worth.12
B.
Other Relevant Fair Lending and Anti-Discrimination Statutes
I briefly describe three other statutes that have been important to the development of fair
lending law and anti-discrimination law generally.
First, the Community Reinvestment Act of 1977 13 is a prominent fair lending law that “is
intended to encourage depository institutions to help meet the credit needs of the communities
in which they operate.”14 The CRA was enacted to correct the problem of “redlining,” which
“refers to the illegal practice of refusing to make residential loans or imposing more onerous
terms on any loans made because of the predominant race, national origin etc., of the residents
of the neighborhood in which the property is located.” 15 The CRA today is notably applied
during bank examinations conducted by the bank regulatory agencies.16
Second, the Fair Housing Act of 1968 has provisions that are roughly parallel with the ECOA, as
applied to housing rather than the extension of credit.17 The FHA prohibits discrimination in
housing “because of race, color, religion, sex, familial status, or national origin.”18 The FHA
explicitly governs discrimination in advertising, making it unlawful “[t]o make, print, or publish, or
cause to be made, printed, or published, any notice, statement, or advertisement, with respect
to the sale or rental of a dwelling that indicates any preference, limitation, or discrimination
based on race, color, religion, sex, handicap, familiar status, or national origin.”19
Third, Title VII of the Civil Rights Act of 1964 is the well-known law that prohibits discrimination
in employment.20 It is an unlawful employment practice for any employer “to fail or refuse to hire
or to discharge any individual, or otherwise to discriminate against any individual with respect to
his compensation, terms, conditions, or privileges of employment, because of such individual’s
race, color, religion, sex, or national origin.”21 Title VII makes it unlawful “to print or publish or
cause to be printed or published any notice or advertisement relating to employment” where
there is an indication of “any preference, limitation, specification, or discrimination, based on
race, color, religion, sex, or national origin.”22 The relevance of Title VII to fair lending and fair
marketing goes beyond advertising for employment, because Title VII cases have developed
much of the jurisprudence for anti-discrimination law generally, including in the area of disparate
impact discussed below.23
C.
The Central Role of Disparate Impact for Enforcement of the ECOA
Fair Lending in the 1990’s. Although the ECOA was enacted in 1974, enforcement of the
broad anti-discrimination provisions in the ECOA began in earnest during the 1990’s. In 1992,
the Federal Reserve of Boston released a detailed study on mortgage loan denial rates in the
Boston area, examining whether racial and ethnic disparities in mortgage loan denial rates
reflected evenhanded use of legitimate credit standards. The Boston Fed study, as it became
widely known, used Home Mortgage Disclosure Act (HMDA) data that showed “patterns of
relatively low mortgage lending in minority areas and indicated that black and Hispanic
applicants were denied loans two to three times as often as whites.”24
In 1994, the federal banking agencies issued the Interagency Policy Statement on
Discrimination in Lending. 25
The agencies used strong, clear language in stating:
“Discrimination in lending on the basis of race or other prohibited factors is destructive, morally
2
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
repugnant, and against the law.”26 The agencies discussed three methods of proof of lending
discrimination that courts had found under the ECOA and the Fair Housing Act:

“Overt evidence of discrimination,” when a lender blatantly discriminates on a
prohibited basis;

Evidence of “disparate treatment,” when a lender treats applicants differently
based on one of the prohibited factors; and

Evidence of “disparate impact,” when a lender applies a practice uniformly to all
applicants but the practice has a discriminatory effect on a prohibited basis and is
not justified by business necessity.”
As discussed further below, the most relevant among these theories for fair marketing is likely to
be the disparate impact theory, where facially neutral practices have a discriminatory effect on a
protected class.
Along with this policy statement, the Department of Justice and federal banking agencies
stepped up their fair lending enforcement during the 1990’s, with multiple lawsuits and resulting
consent decrees based on a “pattern or practice” of lending discrimination.27 The agencies in
1995 issued significant changes to the CRA regulations, which gave more detailed requirements
about the lending, investment, and services that a financial institution provided to communities
in its service area.28 Fair lending also became a more prominent part of bank examinations,
with banking agencies offering guidance on how to comply with fair lending laws, 29 and
emphasizing that fair lending compliance programs would be scrutinized during bank
examinations.30
The Subprime Crisis, Renewed Attention to Fair Lending, and Re-affirmation of the
Disparate Impact Test. From the end of the 1990’s until the mortgage defaults that began in
2006 and 2007, there was relatively little new policy development in the fair lending area. Fair
lending issues have become much more prominent since the financial crisis hit, triggered by
concerns about disproportionate subprime lending to minorities as well as actions by the new
Consumer Financial Protection Bureau (CFPB).
Encouraging homeownership was a domestic policy priority during the 1990's and early
2000's,31 and subprime loans were viewed by many as a way to increase homeownership. 32 To
compensate for increased credit risk, subprime loans carry higher interest rates and were
generally excluded from being pooled into mortgage-based securities by Fannie Mae and
Freddie Mac.33
Minority communities have been disproportionately affected by subprime lending. A 2000 report
by the Department of Housing and Urban Development looked at subprime lending between
1993 and 1998 and found that subprime loans were five times more likely in black
neighborhoods than in white neighborhoods. 34 While much of this disparity may track with
differences in income levels among those neighborhoods, the study also found that
homeowners in high-income black areas were twice as likely as homeowners in low-income
white areas to have subprime loans.35 A May 2002 study by the Center for Community Change
concluded that significant racial disparities existed in subprime lending, and that these
disparities actually increased as borrower income increased.36 In 2009, HUD evaluated HMDA
data in nine major metropolitan areas, finding strong associations between African-American
borrowers and the probability of obtaining higher-priced loans, even when accounting for credit
3
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
score information.37 HUD cautioned that mortgage market discrimination was "coming home to
roost" by undermining the stability and vitality of minority neighborhoods during the height of the
subprime crisis.38
The CFPB was created by the Dodd-Frank Act of 2010, as part of the response to the financial
crisis. It has broad jurisdiction over consumer financial issues including fair lending. The CFPB
took over from the Federal Reserve the lead role in issuing Regulation B to implement the
ECOA, 39 and has issued detailed guidance on its fair lending efforts.
In 2012, the CFPB issued a Bulletin that reviewed applicable legal sources and concluded that
the disparate impact test “would be one method of proving lending discrimination under the
ECOA and Regulation B.”40 The CFPB specifically concurred with the 1994 Interagency Policy
Statement, which recognized three methods for proving lending discrimination under the ECOA:
overt discrimination, disparate treatment, and disparate impact. As explained in Regulation B,
the legislative history indicated that Congress intended the ECOA to use the disparate impact
test as outlined in the employment field in cases such as Griggs v. Duke Power Co.41 The CFPB
reiterated this language from Regulation B:
The act and regulation may prohibit a creditor practice that is discriminatory in
effect because it has a disproportionately negative impact on a prohibited basis,
even though the creditor has no intent to discriminate and the practice appears
neutral on its face, unless the creditor practice meets a legitimate business need
that cannot reasonably be achieved as well by means that are less disparate in
their impact.42
II.
Comparing Fair Lending and Fair Marketing
I believe this background on fair lending and other anti-discrimination statutes aids in assessing
discrimination issues for online marketing. This Part of the White Paper examines reasons why
fair lending approaches may or may not be analogous to fair marketing. It concludes there is no
categorical reason why online marketing should be exempt from anti-discrimination principles,
although there may be significant factual differences to understand in applying such differences.
Among possible distinctions, I examine five:
1. Financial services are heavily regulated and supervised, in contrast to online marketing.
2. The financial services industry is dominated by large companies, in contrast to the many
start-ups and small businesses engaged in online marketing.
3. The decision by the Congress to enact the ECOA and other specific statutes means that
other areas should not be regulated as heavily.
4. One justification for existing anti-discrimination statutes is to address a history of
government action in lending and housing that perpetuated discrimination for many
years. No similar history of government action exists for online marketing.
5. Existing statutes target especially important aspects of consumer participation in the
market, in contrast to the lower stakes for much of online marketing.
Although each of these distinctions is potentially important, there are of course important
counter-arguments to each of them. First, it is true that financial services firms have a long
4
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
history of government regulation and supervision, notably for banks insured by the FDIC, where
taxpayer funds have historically been at risk in case of bank failures. With that said, the ECOA
applies far outside the scope of companies historically subject to regular supervision. It applies
to all creditors, including “any person who regularly extends, renews, or continues credit.”
Similarly, the Fair Housing Act and Title VII apply broadly, and not only to regulated financial
firms.
Second, the existence of large firms does not effectively distinguish fair lending from fair
marketing. Large firms are prominent and have high market shares in commercial banking,
investment banking, insurance, and other financial markets. These large firms generally have a
greater ability than small firms to develop sophisticated fair lending compliance programs. Large
firms also exist in online advertising, however, including some of the firms with the largest
market capitalization in the world. Small firms also exist in lending, and notable fair lending
enforcement actions have applied to small community banks. Because both large and small
firms exist in both financial services and online marketing, a key policy goal would seem in both
settings to be how to make compliance scalable – appropriate to the size and complexity of the
company.
Third is the issue of what weight to put on the legal difference between sectors where Congress
has enacted specific anti-discrimination laws, such as lending, housing, and employment, and
other areas where no such laws exist. As a matter of democratic principle, decisions by the
elected representatives about where to regulate deserve an important degree of deference.
With that said, the anti-discrimination principles embodied in those statutes are held in our
society far more generally than just for a few limited sectors. If there is proof of overt or
intentional discrimination in how a company conducts its marketing, then I have little doubt that
there would be a swift and strong public reaction against that company. There is no reason to
believe that the Congress or the general public believes online marketing to be an area where
discrimination is permitted.
Fourth, a lack of historical government action encouraging online discrimination does not negate
the importance of applying anti-discrimination statutes to protected classes. In the housing
market, there is a well-documented history of government actions that fostered redlining and
lending discrimination for many years. For instance, Federal Housing Administration manuals in
the 1930’s and afterwards down-rated loans when there was “infiltration of inharmonious racial
or nationality groups,” and they explicitly encouraged segregation by calling racially restrictive
covenants “a favorable condition” for a loan.43 In light of this history, one rationale for the ECOA
and the Fair Housing Act was to remedy the earlier government actions that promoted
discrimination. I are not aware of any similar government actions that encouraged discrimination
in online marketing, and so one rationale for existing statutes does not apply to such marketing.
With that said, however, the scope of anti-discrimination statutes sweeps far beyond specific
remedies for specific past government actions. The anti-discrimination statutes today apply to a
range of protected classes that, as a societal and legislative judgment, should not be treated
worse than others.
Fifth, one might argue that existing anti-discrimination laws apply to a limited set of the
transactions of highest importance to individuals – housing, employment, and mortgages and
other major loans such as for automobiles. On this theory, online marketing may consist of
generally less important transactions that do not merit enforcement of anti-discrimination
principles. In response, I note that existing statutes apply to small transactions as well – small
loans, short-term rentals, and transient employment. As a matter of enforcement discretion, the
CFPB and other agencies may not target such smaller transactions, but there is no
5
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
jurisprudence suggesting that the existing statutes apply only to large transactions. In addition,
the overall volume of online advertising and purchases is large and growing, both in dollar value
and in importance in the overall purchases of many consumers. If patterns of unfair marketing
do exist, then the effects of any such discrimination would apply to this large volume of online
advertising and purchases.
Considering these five possible distinctions, I have not discerned any categorical reason why
online marketing should be exempt from the anti-discrimination principles embodied in the
ECOA, Fair Housing Act, Title VII and numerous other laws. I now turn to some specific topics
where the history of fair lending may inform how to treat fair marketing.
III.
Four Lessons for Fair Marketing from Fair Lending Law
I highlight four lessons from fair lending law for fair marketing. First, I show that fair lending law
has often affirmatively encouraged targeted marketing, especially to reach out to minority or
other communities that otherwise would lack equal access to credit. Second, this positive view
toward targeted marketing operates within important constraints – the marketing should not
have a negative and disparate impact on protected classes. There is thus a paradox in the fair
lending approach to targeted marketing – it may be either encouraged or discouraged but the
law is not clear in what contexts. Third, in light of the fair lending jurisprudence related to
marketing, I highlight how existing statutes already make online marketing illegal where there is
discrimination in lending, housing, or employment. Fourth, I discuss the different sources of data
that may be relevant to a disparate impact claim, including ways in which such claims may
actually be easier to establish for fair marketing than for fair lending. A key issue going forward
will be clarifying when a company would be expected to gather and analyze such data.
1. Fair Lending Law Affirmatively Encourages Targeted Marketing to Protected Classes
In privacy policy involving the FTC, there have been intense debates about the role of targeted
marketing. For instance, the FTC has played a leading role in supporting the Do Not Track
process of the World Wide Web Consortium, and I chaired the W3C Do Not Track process in
2012-13. At the W3C, online advertisers and others strongly supported online behavioral
advertising (OBA), stressing its economic importance and ways in which OBA enables
consumers to find the goods and services they prefer. By contrast, consumer groups and others
questioned the consumer benefits of targeted marketing, and argued that, at a minimum,
consumers should have usable and effective controls to limit such marketing. Similar
disagreements about the value of targeted marketing compared with the value of privacy
protection have occurred in numerous other privacy policy debates.
While there remains considerable disagreement about the value of OBA,44 micro-targeting has
become an increasingly important practice in the online ecosystem. Analytics and Big Data use
vast pools of data in order to develop mechanisms to categorize and organize. Advertisers and
business have used these tools to hone their abilities to target a wide range of different
consumer segments and demographics. Despite skepticism, the result has been a major shift in
the online economy. For example, the theory of the “Long Tail” argues that the online economy
is shifting way from mainstream products and markets at the head of the demand curve towards
a larger number of niche products and interests. 45 Niche products can be matched to niche
interests because micro-targeting can reveal those preferences and connect buyers and sellers
in those markets.
6
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
The role of targeted marketing in fair lending, however, has been noticeably different from the
skepticism of such marketing expressed by some in the privacy debates. In fair lending, many
settlements have included requirements for lenders to do more targeted marketing. For
instance, in 1993, the Department of Justice settled an $11 million case with Chevy Chase
Federal Savings Bank for refusing to market its services in minority neighborhoods. 46 Chevy
Chase agreed to open new mortgage offices in black neighborhoods and take steps to equalize
its market share of mortgage loans with minorities. It also agreed to extensively advertise its
services, target sales calls to real estate professionals in black neighborhoods, and provide
training in affirmative marketing programs. Affirmative marketing requirements continue to have
an important role in fair housing settlements. In 2009, HUD entered into a consent decree with
Westchester County that required the county to allocate $51.6 million to construct and
affirmatively market affordable housing in overwhelmingly white county municipalities to combat
residential segregation.47
Regulators’ fair lending guides similarly emphasize the importance of targeted advertising to
meet fair lending goals. For instance, the Federal Reserve Bank of Chicago, in implementing
Interagency Fair Lending Examination Procedures, stated that examiners should “determine if a
lower level of marketing effort was made toward prohibited basis groups or geographies.” 48
Similarly, the Office of the Comptroller of the Currency, in writing about successful fair lending
efforts, said: “These banks generally engage in targeted radio and newspaper advertising, for
examples when they introduce a new product or at a time of year when their customer base
returns from extended trips to a home country. Banks have also found that advertising in local
establishments, like small grocery stores (bodegas) and barbershops, adds to a perception of
credibility.”49
A major recent enforcement action underscores the importance of targeted marketing by
lenders. In June 2014 the CFPB ordered GE Capital to provide $169 million to about 108,000
borrowers excluded from debt relief offers because of their national origin. GE Capital had two
promotions that allowed credit card customers with delinquent accounts to settle their balances
by paying off a specific portion of their debt. For instance, GE Capital offered a credit of
between $25 and $100 for customers who paid their minimum amount due, for a defined set of
customers: those with balances greater than $700, a credit score below 670, and with minimum
payment due of over $150. The legal violation was that GE Capital did not extend these offers to
any customer who indicated they preferred to communicate in Spanish or had a mailing address
in Puerto Rico. The CFPB found a violation of the ECOA because “Hispanic populations were
unfairly denied the opportunity to benefit from these promotions.”50
In summary, fair lending law expects lenders to use targeted marketing, especially to ensure
that members of protected classes have equal access to credit and housing.
2. Fair Lending Law Applies Disparate Impact Analysis to Targeted Marketing
Although regulators clearly support enhanced marketing to protected classes, they also see
risks in such marketing. As discussed above, much of the concern arose from subprime
mortgage lending, where aggressive marketing efforts to minority communities appears to have
resulted in disproportionately higher-interest loans and more home foreclosures.51
My research thus far has not revealed a comprehensive explanation within fair lending about
when targeted marketing should be favored or disfavored. Based on my research to date, one
answer may lie in the concept of “steering.” Under the Fair Housing Act, racial steering is
defined as “deliberately guiding loan applicants or potential purchasers toward or away from
7
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
certain types of loans or geographic areas because of race.”52 For instance, targeting subprime
loans to African-Americans because of race would violate fair lending laws. As the CFPB
explains: “A creditor may not advertise its credit services and practices in ways that would tend
to encourage some types of borrowers and discourage others on a prohibited basis. In addition,
a creditor may not use prescreening tactics likely to discourage potential applicants on a
prohibited basis.”53
As applied to fair marketing, the experience with fair lending shows an apparent paradox –
targeted advertising to reach minority groups and protected classes is often expected, yet it is
illegal to use targeted marketing to steer products to those groups. In my view, this topic needs
considerably more research and discussion. A prominent feature of online advertising is to use
micro-targeting to reach specific groups that are difficult or impossible to reach offline. Some of
those groups use languages other than English, but offers in some but not other languages can
trigger liability under the GE Capital case.
Given the doctrine of fair lending law, the applicable legal approach is generally to apply the
three steps of the disparate impact test. First, does the practice have a disproportionately
negative impact on a prohibited basis? Second, if so, does the practice meet a legitimate
business need? Third, could that need be achieved as well by means that are less disparate in
their impact?
Application of this test would first focus on what constitutes a “negative impact.” Presumably
advertising about legitimate opportunities for credit relief, such as in the GE Capital case, would
have a positive impact. Marketing of a high-risk loan product would have a negative effect,
although the difference between “positive” and “negative” products is far from self-evident. Next,
concerning a “legitimate business need,” online marketing would likely include a wide range of
goods and services that are tailored for particular protected classes, including race, national
origin, sex, and age. Those sorts of differentiated services are less likely to be commercially
prominent in credit card and other loans, so the legitimacy of differentiated products may be
more justifiable in some online advertising instances. Third, for specialized goods and services,
the company would examine whether that need can be achieved by less discriminatory means.
In conclusion on developing a fair marketing approach to targeted marketing, considerable
research and policy discussion will be needed to address the paradox about targeted marketing
that exists in fair lending.
3. Sectoral Statutes Prohibit Discrimination in Online Marketing for Lending, Housing, and
Employment
As shown by fair lending regulatory guides and enforcement actions, the ECOA clearly governs
targeted marketing in lending. Although enforcement actions to date have not focused on
Internet advertising, it is clear that the statute applies to Internet advertising. There is nothing to
suggest an ECOA exception for online advertising, and it would contradict the antidiscrimination purpose of the statute to have any such exception.
More broadly, a key legal fact about fair marketing is that the ECOA, Fair Housing Act, and Title
VII apply to both online and offline marketing. I thus recap the relevant statutory language:
ECOA: It is “unlawful for any creditor to discriminate against any application, with
respect to any aspect of a credit transaction – (1) on the basis of race, color, religion,
8
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
national origin, sex or marital status, or age (provided the applicant has the capacity to
contract).”54
Fair Housing Act: It is unlawful “[t]o make, print, or publish, or cause to be made,
printed, or published, any notice, statement, or advertisement, with respect to the sale or
rental of a dwelling that indicates any preference, limitation, or discrimination based on
race, color, religion, sex, handicap, familiar status, or national origin.”55
Title VII: It is unlawful “to print or publish or cause to be printed or published any notice
or advertisement relating to employment” where there is an indication of “any
preference, limitation, specification, or discrimination, based on race, color, religion, sex,
or national origin.”56
The online advertising eco-system is famously complex, and I do not attempt here to apply the
three statutes to all of the possibly relevant factual settings. Instead, I make five observations
based on the plain language of the statutes and interpretation to date.
First, the quoted language in ECOA applies specifically to “creditors” with respect to “any aspect
of a credit transaction.” This language appears to apply in a straightforward way to lenders
when they place online advertisements. Privacy experts are familiar with sectoral regulation,
such as for HIPAA, Gramm-Leach-Bliley, and the Children’s Online Privacy Protection Act. One
implication of this White Paper is that sectoral regulation exists concerning online advertising as
well.
Second, the ECOA applies to brokers as well as “any person who regularly arranges for the
extension, renewal, or continuation of credit.” The language about “regularly arranges” likely will
be important to defining the effect of the ECOA on non-lenders involved in advertisement on
behalf of lenders. Enforcement actions to date have shown an expansive interpretation of who
“regularly arranges” for credit. For instance, car dealerships and car salespersons are covered,
when they help a purchaser get an automobile loan, even when the lender is a non-affiliated
corporation.57
The statutory question, as applied to fair marketing, will be to determine which parts, if any, of
the eco-system are “regularly arranging” for the extension of credit. One chart of the “display
advertising technology landscape” contains numerous niches, such as: advertising agencies;
media-buying platforms; demand-side platforms; ad exchanges; advertising networks; and
publisher tools. 58 As a hypothetical, suppose that there were disparate treatment or disparate
effect at one or more of those advertising-related entities. While it is clear that advertising and
marketing in general are covered by the ECOA, what facts would be needed to establish the
“regular arrangement” of credit by the various actors? Where such facts exist, then the ECOA
would apply.
Third, similar to the ECOA, the Fair Housing Act applies to those who purchase advertisements,
such as a landlord renting an apartment. Online advertisements for apartments are covered the
same as offline advertisements. One important statutory issue, which is a subject for future
research, is what would meet the statutory requirement that the advertisement “indicates any
preference, limitation, or discrimination” concerning a protected class. For online advertising,
this issue will be important for housing advertisement campaigns that narrowly target a specific
population. For instance, it will be important to clarify whether and when the Act covers
advertisement purchasing decisions that will reach members of a protected class far more or
less often than other demographic groups.
9
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
Fourth, similar to the ECOA, it is important to determine what actors beyond the landlord or
other advertising purchaser are covered by the Fair Housing Act. The statute applies to those
who “make, print, or publish” the advertisements. This language would seem to apply, for
instance, to online publishers as well as those who “make” the advertisement.59 Even more
vaguely, the Act applies to those who “cause to be published” a prohibited advertisement. Once
again, it is far from clear who in the display advertising technology landscape plays this causal
role.
Fifth, the advertising language in Title VII is quite similar to the Fair Housing Act, and additional
research is needed to clarify the scope of coverage for online employment advertisements.
In summary, it is surprising how little of the Big Data and privacy debates to date have focused
on the existing statutes that apply to online marketing. Sectoral regulation exists for
discrimination in online marketing similar to how sectoral regulation exists in the U.S. for many
privacy issues.
4. Data Relevant to a Disparate Impact Analysis May Often Be More Available for Online
Marketing Than For Traditional Fair Lending Cases
The history of fair lending enforcement is intertwined with the availability of data to detect and
prove possible violations of anti-discrimination laws. Much of the fair lending enforcement in the
1990’s relied on data generated under the requirements of the Home Mortgage Disclosure Act
(HMDA). That law requires financial institutions to report certain data about applications and
approval for home loans, including the ethnicity, race, sex, and income of any applicant.
Borrowers provide demographic information in the course of applying for a mortgage.
In fair lending enforcement, a first step has often been to examine an institution more carefully if
there is an apparent discrepancy based on race in mortgage loans. Where such a discrepancy
exists, the next step is to do more careful regression analysis, to determine whether factors
other than membership in a protected class explain the discrepancy. 60 If a significant
discrepancy still exists, then that can be evidence of discrimination in violation of the ECOA.
One recurring issue in fair lending enforcement is that the lenders, outside of the mortgage
context, often have no good demographic data about which potential borrowers belong to a
protected class. It is not a standard part of a credit card application, for instance, for lenders to
ask about race and national origin, and any lender who asked customers for that information
could face a negative customer reaction and bad press. For this reason, the General
Accountability Office reported in 2009 that fair lending oversight activities on areas such as
automobile, credit card, and business lending have been considerably more limited.61 In recent
years, the CFPB has made a higher priority of automobile and credit card loans, but data issues
persist. The CFPB is currently engaged in a procedure to clarify what data should be
considered, in what way, for identifying situations that merit further investigation under the
disparate impact approach.62
These data limitations may be easier to address in many online advertising situations, which are
characterized by precisely the Big Data that is the subject of this FTC Workshop. The online
advertising ecosystem generates many data points about an individual or device, and far more
data points than are typically available in the traditional lending relationship. Micro-targeting
routinely estimates the age and gender of those receiving the advertisements. Where microtargeting uses dozens or hundreds of market segments, the relative abundance of data may
10
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
provide useful insights into likely membership into demographic groups such as race or national
origin.
It is beyond the scope of this White Paper to set forth how this relatively abundant demographic
data might be used to assess online lending advertisements or fair marketing generally. In light
of this often-abundant data, an important issue may become whether there are circumstances in
which a company may be legally or ethically expected to analyze such data to detect possible
violations of anti-discrimination law or principles.
The point here is a limited one – data limitations have been an important constraint on fair
lending enforcement to date, but the Big Data available through online advertising quite possibly
means such data limitations will be less important for fair marketing.
Conclusion
Determining the precise scope and contours of fair marketing is an ambitious project. To
alleviate concerns in fair lending, lenders have developed significant compliance programs
including extensive training in fair lending for their personnel. Fair lending law provides decades
of lessons from jurisprudence, regulatory guidance, and industry initiatives that could help guide
next steps for both industry and the FTC to evaluate marketing data and ensure advertising is
done in ways that benefit consumers.
Big Data is changing online marketing, but precedents and lessons exist on how to conduct
marketing in ways that are lawful, fair, and consistent with society’s commitment to antidiscrimination principles. I look forward to working with stakeholders to develop the idea of fair
marketing, and appreciate the Federal Trade Commission’s interest in ensuring that Big Data is
used as a tool for inclusion.
1
Peter P. Swire, Equality of Opportunity and Investment in Creditworthiness, 143 U. PA. L. REV. 1533 (1995); Peter P. Swire, The
Persistent Problem of Lending Discrimination: A Law and Economics Analysis, 73 TEX. L. REV. 787 (1995); Peter P. Swire, Safe
Harbors and a Proposal to Improve the Community Reinvestment Act, 79 VA. L. REV. 349 (1993).
2
15 U.S.C. § 1691(a).
3
15 U.S.C. § 1691(b).
4
12 CFR Part 1002 et seq. Current CFPB guidance for the ECOA is available at CFPB Consumer Laws and Regulations: Equal
Credit Opportunity Act (ECOA), http://files.consumerfinance.gov/f/201306_cfpb_laws-and-regulations_ecoa-combined-june-2013.pdf
(June 2013).
5
15 U.S.C. § 1691a.
6
15 U.S.C. § 1691c.
7
15 U.S.C. § 1691c(c).
8
Id.
9
Id.
10
Id.
11
15 U.S.C. § 1691e(h).
12
15 U.S.C. § 1691e(a) & (b).
13
12 U.S.C. § 2901 et seq.
14
Federal Financial Institutions Examination Council, “Community Reinvestment Act,” http://www.ffiec.gov/cra.
15
Interagency Policy Statement on Discrimination in Lending, Fed. Reg. Apr. 15, 1994. http://www.occ.treas.gov/newsissuances/federal-register/94fr9214.pdf. The term “redlining” derives from lenders drawing a red line around minority neighborhoods
11
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
and not lending into those neighborhoods. The history of empirical evidence related to redlining is discussed in Peter P. Swire, The
Persistent Problem of Lending Discrimination: A Law and Economics Analysis, 73 TEX. L. REV. 787, 806-14 (1995).
16
Federal Financial Institutions Examination Council, CRA Examinations, http://www.ffiec.gov/cra/examinations.htm.
17
42 U.S.C. § 3601-3631.
18
42 U.S.C. § 3604.
19
42 U.S.C. § 3604(c).
20
42 U.S.C. § 2000e.
21
42 U.S.C. § 2000e-2(a).
22
42 U.S.C. § 2000e-3. There is an exception providing that “such a notice or advertisement may indicate a preference, limitation,
specification, or discrimination based on religion, sex, or national origin when religion, sex, or national origin is a bona fide
occupational qualification for employment.” Id.
23
For a discussion of implications of Title VII for big data and privacy, see Solon Barocas & Andrew D. Selbst, Big Data’s Disparate
Impact, www.ssrn.com/2474789.
24
Interagency
Policy
Statement
on
Fair
Mortgage
http://www.federalreserve.gov/bankinforeg/interagencystatement.htm.
Lending
Practices,
Oct.
9,
1992,
25
Interagency Policy Statement on Discrimination in Lending, Fed. Reg. Apr. 15, 1994. http://www.occ.treas.gov/newsissuances/federal-register/94fr9214.pdf.
26
Id.
27
UNITED STATES GENERAL ACCOUNTING OFFICE, FAIR LENDING: FEDERAL OVERSIGHT AND ENFORCEMENT IMPROVED BUT SOME
CHALLENGES REMAIN (1996), http://www.gao.gov/products/GGD-96-145.
28
Community Reinvestment Act Regulations, 60 Fed. Reg. 22,156, 22,157 (May 4, 1995) (codified at 12 C.F.R. pts. 25, 203, 228,
345, 563e). See Raymond H. Brescia, Part of the Disease or Part of the Cure: The Financial Crisis and the Community
Reinvestment Act, 60 S.C. L. REV. 617 (2009).
29
FEDERAL
DEPOSIT
INSURANCE
CORPORATION,
https://www.fdic.gov/regulations/resources/side/.
SIDE
BY
SIDE:
A
GUIDE
TO
FAIR
LENDING
(1996),
30
FEDERAL FINANCIAL INSTITUTIONS EXAMINATION COUNCIL, INTERAGENCY FAIR LENDING EXAMINATION PROCEDURES (Jan. 22, 1999),
https://www.fdic.gov/news/news/inactivefinancial/1999/FIL9914b.html.
31
UNITED STATES DEP'T OF HOUSING & URBAN DEVELOPMENT, THE NATIONAL HOMEOWNERSHIP STRATEGY: PARTNERS IN THE AMERICAN
DREAM (1995), available at http://confoundedinterest.files.wordpress.com/2013/01/nhsdream2.pdf; Press Release, Fact Sheet:
America's Ownership Society: Expanding Opportunities, The White House (Aug. 9, 2004), http://georgewbushwhitehouse.archives.gov/news/releases/2004/08/20040809-9.html.
32
See KATHLEEN C. ENGEL & PATRICIA A. MCCOY, THE SUBPRIME VIRUS: RECKLESS CREDIT, REGULATORY FAILURES, AND NEXT STEPS
21 (2011) (citing Julie Williams, Chief Counsel of the Office of the Comptroller of Currency, lauding “the rise of the subprime
segment . . . in advancing homeownership, especially for minority Americans.”).
33
CONGRESSIONAL BUDGET OFFICE, FANNIE MAE, FREDDIE MAC, AND THE FEDERAL ROLE IN THE SECONDARY MORTGAGE MARKET 6
(2010), http://www.cbo.gov/sites/default/files/cbofiles/ftpdocs/120xx/doc12032/12-23-fanniefreddie.pdf.
34
UNITED STATES DEP’T OF HOUSING & URBAN DEVELOPMENT, UNEQUAL BURDEN: INCOME AND
LENDING IN AMERICA (2000), http://www.huduser.org/portal/publications/fairhsg/unequal.html.
35
RACIAL DISPARITIES IN SUBPRIME
Id.
36
CALVIN BRADFORD, RISK OR RACE? RACIAL DISPARITIES AND THE SUBPRIME MARKET, CTR. FOR COMMUNITY CHANGE (2002), available
at http://www.knowledgeplex.org/showdoc.html?id=1032.
37
UNITED STATES DEP’T OF HOUSING AND URBAN DEVELOPMENT, RISK OR RACE: AN ASSESSMENT OF SUBPRIME LENDING PATTERNS IN
NINE METROPOLITAN AREAS (2009), http://www.huduser.org/portal/publications/pdf/risk_race_2011.pdf.
38
Id. at 47-48.
39
Regulation
B
is
codified
at
12
CFR
Part
idx?SID=1716fb3f1e1fb0685459a237478ca396&node=pt12.8.1002&rgn=div5.
40
CFPB
Bulletin
2012-044
(Fair
http://files.consumerfinance.gov/f/201404_cfpb_bulletin_lending_discrimination.pdf.
41
401 U.S. 424 (1971).
42
12 CFR part 1002, Supp. I, § 1002.6, ¶ 6(a)-2.
12
1002,
Lending)
http://www.ecfr.gov/cgi-bin/text(Apr.
2012),
LESSONS FROM FAIR LENDING LAW FOR FAIR MARKETING AND BIG DATA
43
Peter P. Swire, The Persistent Problem of Lending Discrimination: A Law and Economics Analysis, 73 TEX. L. REV. 787, 794
(1995).
44
See Jason Kint, Behavioral Advertising Might Not Be As Crucial As You Think, ADAGE
http://adage.com/article/datadriven-marketing/behavioral-advertising-crucial/291858/.
45
Chris
Anderson,
Long
Tail
101,
THE
LONG
http://www.longtail.com/the_long_tail/2005/09/long_tail_101.html.
TAIL:
CHRIS
ANDERSON'S
BLOG
(Feb. 25,
(Sep.
08,
2014),
2005),
46
Complaint, United States v. Chevy Chase Fed. Sav. Bank, Civ. Action No. CV94-1824JG (D.D.C. 1994); see also National Fair
Housing Advocate Online, Justice Department: D.C. Bank Pays $11 Million For Failing to Service Predominantly Black Areas,
http://www.fairhousing.com/index.cfm/index.cfm?method=page.display&pagename=September-October_1994_Page3 (last visited
Sep. 6, 2014).
47
Stipulation and Order of Settlement and Dismissal, U.S. ex rel. Anti-Discrimination Center of Metro New York Incorporated v.
Westchester County, No. 1:06-cv-2860-DLC (S.D.N.Y. Aug. 10, 2009), http://www.hud.gov/content/releases/settlementwestchester.pdf.
48
FEDERAL RESERVE BANK OF CHICAGO, BANKER’S GUIDE TO RISK-BASED FAIR LENDING EXAMINATIONS 3,
https://www.chicagofed.org/digital_assets/others/utilities/about_us/cca/bankers_guide_to_risk_based_fair_lending_examinations.pdf
.
49
OFFICE OF COMPTROLLER OF THE CURRENCY, REACHING MINORITY MARKETS: COMMUNITY BANK STRATEGIES 10 (2006),
http://www.occ.gov/topics/community-affairs/publications/insights/insights-reaching-minority-markets.pdf.
50
Press Release, CFPB Orders GE Capital to Pay $225 Million in Consumer Relief for Deceptive and Discriminatory Credit Card
Practices, Consumer Protection Financial Bureau (June 19, 2014), http://www.consumerfinance.gov/newsroom/cfpb-orders-gecapital-to-pay-225-million-in-consumer-relief-for-deceptive-and-discriminatory-credit-card-practices/.
51
See, e.g., UNITED STATES DEP’T OF HOUSING & URBAN DEVELOPMENT, UNEQUAL BURDEN: INCOME AND RACIAL DISPARITIES IN
SUBPRIME LENDING IN AMERICA (2000), http://www.huduser.org/portal/publications/fairhsg/unequal.html; CALVIN BRADFORD, RISK OR
RACE? RACIAL DISPARITIES AND THE SUBPRIME MARKET, CTR. FOR COMMUNITY CHANGE (2002), available at
http://www.knowledgeplex.org/showdoc.html?id=1032; Jacob W. Faber, Racial Dynamics of Subprime Mortgage Lending at the
Peak, 23 HOUSING POLICY DEBATE 328 (2013) (evaluating 2006 HMDA data to conclude that wealthier black and Latino applicants
were 2.4 times more likely to receive subprime loans than whites.).
52
Federal
Fair
Lending
Regulations
and
http://www.federalreserve.gov/boarddocs/supmanual/cch/fair_lend_fhact.pdf.
Statutes,
53
“Fair
Consumer Financial Protection Bureau, CFPB Consumer Laws and Regulations (ECOA)
http://files.consumerfinance.gov/f/201306_cfpb_laws-and-regulations_ecoa-combined-june-2013.pdf.
54
15 U.S.C. § 1691(a).
55
42 U.S.C. § 3604(c).
Housing
3
(June
Act,”
2013),
56
42 U.S.C. § 2000e-3. There is an exception providing that “such a notice or advertisement may indicate a preference, limitation,
specification, or discrimination based on religion, sex, or national origin when religion, sex, or national origin is a bona fide
occupational qualification for employment.” Id.
57
See, e.g., CFPB Bulletin 2013-02, Indirect Auto Lending and Compliance with the Equal Credit Opportunity Act (Mar. 21, 2013),
http://files.consumerfinance.gov/f/201303_cfpb_march_-Auto-Finance-Bulletin.pdf.
58
Luma
Partners,
Display
Advertising
Technology
Landscape
content/uploads/2010/09/LUMA-Display-Ad-Tech-Landscape-for-AdExchanger.jpg.
(2010),
http://www.adexchanger.com/wp-
59
I note that there is some interplay between anti-discrimination laws and the Communications Decency Act. See, e.g., Chicago
Lawyers' Comm. for Civil Rights Under Law, Inc. v. Craigslist, Inc., 519 F.3d 666 (7th Cir. 2008), as amended (May 2, 2008) (finding
Craigslist not liable for discriminatory postings and user classified ads).
60
Stuart Rossman, Analyzing Disparate Impact Credit Discrimination in the Subprime Lending Market, 1533 PLI/CORP 601, 608-10
(2006).
61
GENERAL ACCOUNTING OFFICE, FAIR LENDING: DATA LIMITATIONS AND THE FRAGMENTED U.S. FINANCIAL REGULATORY STRUCTURE
CHALLENGE FEDERAL OVERSIGHT AND ENFORCEMENT EFFORTS (2009), http://www.gao.gov/products/GAO-09-704.
62
For discussion, see NATIONAL COMMUNITY REINVESTMENT COALITION, WHITE PAPER: SMALL BUSINESS LOAN DATA:
RECOMMENDATIONS TO THE CONSUMER FINANCIAL PROTECTION BUREAU IMPLEMENTING SECTION 1071 OF THE DODD-FRANK W ALL
STREET
REFORM
AND
CONSUMER
PROTECTION
ACT
OF
2010
(Aug.
7,
2014),
available
at
http://www.ncrc.org/images/PDFs/recommendations-to-cfpb-on-small-business-loan-data.pdf.
13